More settings (H–Q)
Plain-language reference for Omniscio settings whose keys begin with the letters H through Q, this page covering that span of the alphabetical settings reference.
On this page
- How you heard about us
- Activity nudges
- Feature tips
- One-time tips
- Hide asides by default
- Hide repeated auto-wait turns
- Hide routine updates from the inbox
- Project-aware welcome headlines
- Hotkey training mode
- Hub Overseer — what the seat is for
- Hub Overseer — standing instructions
- Image bridge mode
- Hubs in your Inbox
- Model for the first session title
- Instant Fix hotkey
- Instant Fix hotkey on/off
- Instant Fix cleanup level
- Instant Fix writing style
- Interruptions avoided card
- Omni Briefing — auto-play on open
- Omni Briefing — enabled
- Omni Briefing — hotkey
- Omni Briefing — hotkey enabled
- Omni Briefing — model
- JLS Image Studio app panel
- JLS Image Studio MCP integration
- Let Running jobs stop processes
- Let the AI Coaching coach see my recent entries
- Send guided writing to an AI for replies
- Your AI service
- Reply style
- Keep long sessions on their current account
- Key reminders prompt
- Kimi balance monitor
- Kimi balance alert threshold
- KMS summary model
- Landing v2
- Launch on startup
- Marketplace
- Marketplace review queue
- Model for master-debt fixers
- Mission Control sidebar categories
- Meeting transcription service
- Meeting summary model
- MemPalace
- Underline menu letters
- Mind map model
- Mind Map AI daily limit
- Mission Control
- Show text on the collapsed Omni panel
- Mobile device location
- Mobile fast load
- Mobile font size
- Mobile header style
- Mobile instant shell
- Let plugins open on your phone
- Mobile reliability alerts
- Alert me if push stops reaching my phone
- Mobile row density
- Automatically use the newest model
- Keep working when one model runs out
- Narration mode
- Narration prompt override
- Narration read view
- Narration voice reply auto-send
- Enable Vault (KMS)
- KMS Vault Overview
- Vault visual theme
- Vault colour mode
- Vault folder
- Keep only pinned notes as tabs
- Vault summary daily cost cap
- Vault bulk-seed per-run cost cap
- Vault image analysis cost cap
- Refresh stale summaries automatically
- How often to refresh a summary
- Wait before refreshing the same note again
- Expose Vault to spawned agents
- Vault editor font
- Vault editor text size
- Vault editor line width
- Vault editor page colour
- Vault sidebar side
- AI Knowledge Digest
- Daily digest
- Run the daily digest at
- Weekly digest
- Run the weekly digest on
- AI Knowledge Digest per-run cost cap
- Deep-dive session limit
- Digest sections
- Session finished sound
- OpenAI admin key (vendor key watch)
- OpenCode model
- Who pays for each provider
- Outbound webhooks
- Overseer groups
- Paste rich text as Markdown
- Hand agent messages over mid-turn
- Skip redundant read-permission requests
- Agent persona
- Pi model
- Session tool picker tools
- Pika voice
- Plugin auto-check updates
- Lock down plugins' AI agents
- Plugins whose AI agents have your full access
- Background AI sessions a day
- Which boards appear on the calendar
- PM calendar overlay
- Per-board inbox preferences
- Mission Control Inbox level
- Tell me when an automation fails
- Tell me when an automation changes something that involves me
- Ask before opening Omni
- PM inbox digest mode
- PM inbox
- PM inbox quiet hours
- PM inbox quiet hours end
- PM inbox quiet hours start
- PM inbox quiet hours time zone
- Omni's memory of you
- Send nudges right away
- PM setup preset
- AI comfort level
- Add my out-of-office from Google Calendar
- Ask automatically when I open the home page
- Use a faster, cheaper AI for What Next
- Pomodoro — auto-enable Focus Mode
- Pomodoro — bring to foreground on phase change
- Pomodoro schedule
- Post-send navigation
- Skip inbox cards for my own pull requests
- PR Merge Queue — auto-merge poll interval
- PR Merge Queue
- PR Merge Queue — shard count
- PR Merge Queue — sharding
- PR Merge Queue — shard index
- Let agents restart your programs so they survive quitting
- When I switch projects
- Provider config sync targets
- Quick Chat per-target hotkeys
- Quick Email saved recipients
- Quick Email signature
- Quick Email undo window
- Quick Launch — follow cursor
- Quick Launch hotkey
- Quick Launch — preserve draft
- Quick Launch — show engine picker
- Quiet hours
- Source checkout folder
- Board density
- Count my week as business days only
- Name your assistant
- Match my Omniscio theme
- Let Omni read your documents, spreadsheets and forms
- Sync to every enabled AI tool
- Run this many PR sessions at once
- Pull requests per session
- Move a hub to another workspace
Plain-language reference for Omniscio settings whose keys begin with the letters H through Q.
How you heard about us
Records the acquisition channel you selected during initial setup ("how did you hear about Omniscio?"). The value is one of the canonical marketing-source options shown in the Setup wizard. It is written once, synced to your cloud profile, and then left alone — you will not be prompted again. This field feeds aggregate marketing reporting; it does not affect any app behavior. If you did not complete the onboarding step it will be absent.
Activity nudges
The cards and hints that react to what you are already doing: saving a phrase you keep typing as a quick reply, removing a quick reply you have stopped using, the reminder that your inbox has piled up, the one-tap (Zap) shortcut coach, the in-session "you do not need to watch this agent" hint, the weekly watch-time recap, and reflection prompts. On by default, and off only when you switch it off yourself; an upgrade never turns it back on after you turn it off. The individual switches underneath still apply, so a nudge appears only when this switch and its own control are both on. Real alerts are not affected.
Feature tips
Short "did you know" cards, one per feature you have not tried yet, plus the shortcut coaching tips that appear while you work. Off by default and opt-in; each feature keeps its own on/off control underneath. Real alerts are not affected.
One-time tips
The cards that appear a single time and then stay away — emptying your inbox by archiving, and turning off desktop notifications. Off by default and opt-in. Real alerts are not affected.
Hide asides by default
Controls how "aside" conversations (side-questions spawned with Ctrl+B that fork a session) are rendered in the transcript. When on, asides start collapsed to a header chip and you click to expand them. When off (the default), asides render fully expanded inline alongside the main session content. Turn this on if you use asides frequently and find them visually noisy in their expanded form.
Hide repeated auto-wait turns
When enabled, collapses repeated "Auto-waited" turns — the prose-wait and scheduled-wakeup lifecycle rows a session emits while it silently waits for you — into a single expandable one-line marker, keeping only the most recent one visible. Only a message you send starts a new series — a message from another agent or a scheduled wake-up does not, so a coordinating session's routine check-ins fold too. The very first wait turn in a series is never hidden, nor is a turn that carries a user message or one that is still in flight. This is a pure display preference and does not alter what is saved. Off by default; opt in from Settings if your chat views feel cluttered by wait rows. Pairs with the fold-auto-wait-notices option.
Hide routine updates from the inbox
Keeps routine progress reports out of your inbox. When an Overseer, or a session that belongs to a crew, finishes a turn and says its own reply was only a routine update, that turn does not appear in your inbox, the sidebar's Needs You list or the badge count. Nothing is deleted.
A turn that asks you something is never hidden, and neither is a turn the session addressed to you, or one that arrives while the session was already waiting on you. Ordinary sessions you start yourself still reach your inbox as usual. Turn this off and every finished turn lands in the inbox the way it always did. Find it in Settings > Sessions. Default: on.
Project-aware welcome headlines
Controls the big welcome heading on a brand-new session. On (the default), the heading rotates through a built-in set of friendly lines, and in a project that has some history it sometimes shows a short headline written for that project. To write those, Omniscio occasionally asks a low-cost AI model to read your recent session titles and opening messages in the project, then keeps the result for a day or more. Each session picks its heading once and keeps it, so it never changes while you look at it. Off, nothing is sent and you only ever see the built-in set.
Hotkey training mode
An opt-in training aid that blocks the first click on any control that also has a keyboard shortcut, showing a "Press <key>" coaching bubble at your cursor instead. There is a soft escape hatch you can always take: the second click within about three seconds is still blocked and shows "Click again to allow", and the third click performs the action with a green "Allowed" confirmation. This trains you to reach for the keyboard instead of the mouse. An item inside an open menu (a ⋯ menu, a right-click menu, a dropdown) is never blocked and runs on the first click. Keyboard shortcuts themselves are unaffected, and the mode has no effect on touch-only devices. Off by default. Find it in Settings → Sessions.
Hub Overseer — what the seat is for
Sets what the hub-wide Overseer seat is responsible for. everything — the default — covers your own inbox plus every crew and every Overseer. inbox-only screens your inbox and leaves the crews alone, for someone who wants intake handled but wants the crews to run unattended. The seat reads this on its next wake, so a change needs no restart and no re-assignment. Find it in Overseers hub → Fleet settings → How they behave. Accepted values: everything | inbox-only.
Hub Overseer — standing instructions
Your own standing instructions to whoever currently holds the hub-wide Overseer seat, delivered with every wake the holder takes. Free text, up to 2,000 characters, and never parsed by the app — it is passed through as written. Empty by default, meaning the seat brings nothing of yours beyond its own brief. Set it in Overseers hub → Fleet settings → How they behave.
Image bridge mode
Controls whether images you attach to a session are automatically described for text-only AI models. off (the default) leaves images unprocessed; auto sends attached images to a vision-capable model and injects the generated description into the session context. Use auto when you routinely work with text-only providers and need image content to reach the model. Note that auto adds an extra model call per image, so it has a small cost.
Hubs in your Inbox
Lets you keep a whole hub out of your inbox without touching the hub itself. Each hub has three choices: Inbox, Quiet or Hidden. This setting remembers which hubs you chose to hide.
A hidden hub's items leave the Inbox list and the Inbox count, and the empty-inbox message no longer counts them. The hub keeps working, and its row in the sidebar still shows its needs-you badge. You can also hide a hub by right-clicking its group header in the sidebar inbox, and an Undo message appears. Find the full list in Settings > Inbox. Default: no hubs hidden.
Model for the first session title
Which AI model writes a session's title the first time one is generated. Leave it on Automatic and Omniscio picks for you — a fast, low-cost model, which runs on your own machine when the on-device title model is available. Only used when session naming is set to AI titles; the separate refresh-title model handles later re-titling.
Instant Fix hotkey
The keyboard shortcut for Instant Fix (default Ctrl+Shift+Alt+G): highlight text, press it, and an AI fixes it and replaces your selection in place. Rebind it in Settings → Keyboard Shortcuts → System-wide.
Instant Fix hotkey on/off
Whether the Instant Fix hotkey is registered (default on). It also needs AI Text Correction switched on in Settings → Lab. Turn it off to free the shortcut without losing your binding.
Instant Fix cleanup level
How much Instant Fix changes your text: "Fix and tidy up" (default) fixes mistakes and smooths clearly clumsy wording; "Fix mistakes only" fixes spelling, grammar and punctuation and nothing else. Both keep your voice. Found under Settings → Sessions → Per-feature AI models → Instant Fix.
Instant Fix writing style
An optional note (up to 500 characters) describing how you write, sent to the AI with each Instant Fix as style guidance. Empty by default.
Interruptions avoided card
Shows or hides the "Interruptions Avoided" card in the Statistics → Overview panel. The card displays a running tally of moments Omniscio handled something for you autonomously — self-archived items, automatic reply holds, dev-pipeline auto-approvals — so you were not interrupted. The underlying counting always runs (it is near-free), so toggling this only controls whether the card is visible. Default on.
Omni Briefing — auto-play on open
When on, the Omni Session Briefing is automatically spoken aloud the moment you open a session, without you pressing a button. Requires Omni Briefing to be enabled. Off by default; useful for a hands-free workflow where you want to hear a spoken context summary immediately on focus. Combines with jarvisBriefingHotkeyEnabled for manual-trigger use.
Omni Briefing — enabled
Master toggle for Omni Session Briefing, a voice-spoken summary of a session's recent context. When on, opening a session gives you the option to hear a spoken recap of recent activity before you type. Off by default. Requires at least one text-to-speech provider to be configured. See also jarvisBriefingAutoOnOpen and jarvisBriefingHotkey.
Omni Briefing — hotkey
The global keyboard shortcut that fires an Omni Briefing on demand (default Ctrl+Shift+J). Accepts a single Electron accelerator string or a list of alternates. Registered whenever jarvisBriefingHotkeyEnabled is on, whether or not Omni Briefing itself is on. Change it in Settings → Keyboard Shortcuts → System-wide, the Omni briefing row, if it conflicts with another application.
Omni Briefing — hotkey enabled
Whether the Omni Briefing global hotkey is registered with the OS. It is on by default, even while Omni Briefing itself is off, so Ctrl+Shift+J is taken from the first start. Turn this off to free the key while still keeping Omni Briefing available via the in-app button.
Omni Briefing — model
The AI model used to generate the spoken briefing script. Defaults to a lightweight OpenRouter-hosted model. You can change this to any model Omniscio can reach if you prefer a different voice or cost profile. Accepts a model-id string up to 100 characters. Setting → Sessions → Per-feature AI models.
JLS Image Studio app panel
Shows or hides the Image Studio sidebar row, which opens the native Image Studio panel rather than the jls-image-studio.web.app website. Off by default; turn it on with Enable Image Studio in Settings → Built-in Apps. Separate from the Image Studio tool service in Accounts & Providers, whose launch toggle was retired 2026-09-19 (only its API key remains).
JLS Image Studio MCP integration
Retired (2026-09-19). No AMC-built MCP server composes into a session any more, so the Image Studio launch toggle is gone. The service's API key is still configured in Accounts & Providers and is still read by the Image Studio CLI routes.
Let Running jobs stop processes
Decides whether the Running jobs panel is allowed to stop things. The panel always lists the test and build processes Omniscio is tracking, whether this is on or off. This setting only controls the three actions that end a process: Kill, Release slot and Clean up orphans.
When it is off, those buttons are grayed out and the panel tells you that killing and releasing are off. The same actions are also refused if you try them from your phone or the command line. Leave it on if you want to clear a stuck test or build yourself. Turn it off if you only want to look and never stop anything. Find it in Settings > Features. Default: on.
Let the AI Coaching coach see my recent entries
Lets AI Coaching sessions read your most recent Journal entries so the coach knows what has been on your mind. Off: an AI Coaching session is refused at the Journal, so the coach never sees an entry and no approval card is raised for it; other agents keep the ordinary one-click approval. The Journal also learns no new facts from your entries while it is off, and never goes back to learn from an entry finished while it was off. Turning it off deletes nothing. Shown in Settings → Plugins → Journal whenever Journal is on. Changes only on this computer: a paired phone cannot switch it. On/off; default on.
Send guided writing to an AI for replies
Sends what you write in a guided reflection to the service picked in "Your AI service" so it can reply, together with a short background about you: about two weeks of your journal entries and, while AI Coaching is on, your coaching profile and key facts. The Talk with Coach greeting sends the same background. When you finish an entry, one more short request picks out a few key facts to remember, and you can remove any of them. Off: nothing you write leaves this computer, guided messages still save but get no reply, the chat says guided replies are off, the greeting is the fixed welcome, and the service picker is hidden. Turning it off deletes nothing. Shown in Settings → Plugins → Journal whenever Journal is on. Changes only on this computer: a paired phone cannot switch it. On/off; default on.
Your AI service
Picks the one AI service that answers guided Journal reflections and writes the Coach's personal greeting. Your Claude subscription (the default, and the best results) runs a short private Claude reply on your own Claude sign-in, so it adds no per-message cost. Your own AI service uses the AI keys in Settings → My API Keys (OpenRouter, DeepSeek and others, and never the Omniscio allowance) and may cost per message. Journal never switches between the two on its own: if the picked service can't answer, the chat says why and your message stays saved. Shown in Settings → Plugins → Journal whenever Journal is on with guided replies on, and in AI Coaching's settings whenever AI Coaching is on (it is one setting, so changing it in either place changes both). Changes only on this computer: a paired phone cannot change it. Choice; accepted values: claude-subscription | ai-service.
Reply style
Picks how guided Journal replies sound. Calm (the default) is warm and unhurried. Upbeat is encouraging and notices your wins. Minimal is brief and plain, usually one short question. It changes the voice only: the layout, what is sent and which service answers stay the same. The Talk with Coach greeting is not affected. Shown in Settings → Plugins → Journal whenever guided replies are on. Choice; accepted values: calm | upbeat | minimal.
Keep long sessions on their current account
When Omniscio spreads your sessions across several Claude accounts, a session that has just been working with a large conversation is left on its current account instead of being moved. Moving it throws away the saved copy of the conversation the model had already read, so the new account would have to read all of it again, which is slower.
This only affects the tidy-up that moves sessions around to balance accounts. An account that has run out, been signed out or hit a limit is still left right away. It needs "Distribute sessions across accounts" to be turned on, and it is an advanced option. Find it in Settings > Performance. Default: on.
Key reminders prompt
Controls whether a "Key Reminders" block is injected near your message on every session turn. When on (the default), each turn reminds the agent of the most important habits: assume you can and look harder, believe bug reports, act and prove it with evidence, default to yes, fix root causes, keep it simple, and only ask when it is genuinely your call. Turn off to stop injecting the reminder block. Settings → Sessions → Prompting.
Kimi balance monitor
When enabled, Omniscio periodically checks your Kimi (Moonshot) API account balance and raises an inbox alert when it falls below the threshold set in kimiBalanceThresholdUSD. Off by default. Requires a Kimi API key stored in Accounts & Providers. Kill switch: AMC_DISABLE_KIMI_BALANCE_MONITOR.
Kimi balance alert threshold
The USD balance level below which a low-balance inbox alert fires for your Kimi (Moonshot) account. A value of 0 disables the alert entirely. Default is $5. Range: $0–$10,000. Only relevant when kimiBalanceMonitorEnabled is on.
KMS summary model
The AI model used by the Vault's AI features: writing note summaries, and running the AI actions in the note editor (rewrite, shorten, explain, fix grammar, continue and ask). null (the default, labeled "Automatic" in the UI) uses Omniscio's built-in default model. Set a specific model string to override it for both kinds of call; both also count against the Vault AI daily cost cap (nothariSummaryCostCapUsd). Advanced option; find it under Settings → Sessions → Per-feature AI models. Synonyms: knowledge vault summary model, note summary model.
Landing v2
Which landing path the auto-lander uses for a finished branch. On by default — the new landing judge. While it is on, the checks a branch already passed when it was tagged are not run again before it lands, as long as that is on record for the commit being landed; a branch with no such record still takes the full check chain, and the merge itself is always proved sound at land time. Flipping it off is the one-click way back to the previous landing behaviour, and it takes effect on the very next land with no restart and no lost work in flight. Toggle; accepted values: true | false.
Launch on startup
Whether Omniscio starts automatically when you log in to your computer (Electron login item registration). Off by default. Turn on if you want Omniscio running in the background from the moment your machine boots. Toggle in Settings → System.
Marketplace
Shows or hides the Plugin Marketplace virtual project in the sidebar. The Marketplace lets you browse, install, and manage community-built plugins. Off by default for all users (a force-off migration applied 2026-05-26); re-enable in Settings → AI Coach. Turning it off hides the sidebar row but does not uninstall any plugins you have already installed.
Marketplace review queue
Shows or hides the Marketplace Review virtual project — an admin queue for reviewing and approving plugin submissions. Off by default; intended for plugin administrators. If you are a marketplace admin, install Marketplace Review from Marketplace → Inbuilt; this switch is no longer a Settings row. Unrelated to the end-user Marketplace toggle above.
Model for master-debt fixers
Which model the master-debt fixers run on when they clear failures that were already present on the base branch, before your branch's own results are considered. DeepSeek V4.1 Flash is the default — that work is mechanical and highly repeated, so the cheapest capable engine is the right pick. If the engine you choose is not available on this machine, the fixers quietly fall back to the reporting project's own engine, so a fix never stalls on this choice.
Mission Control sidebar categories
Choose which sections appear in the Mission Control sidebar: All Projects, Dashboards, Goals, Portfolios, Documents, Spreadsheets and Forms, one switch each. New users see All Projects; a category switches itself on when you create something in it, and on upgrade the categories you already use are switched on once. Change it here or on the Sidebar card of Mission Control's own Settings page.
Meeting transcription service
Which speech-to-text engine transcribes your meetings. It is deliberately separate from the voice-typing engine, so you can use a different service for each. Defaults to Local — the free on-device engine — so meetings transcribe out of the box with no API key and no chance of a surprise bill. Choose a cloud engine (Deepgram, ElevenLabs, Groq, Grok, Gemini or Meta) if you want higher accuracy or better speaker separation; each of those bills your own key for that provider, which is exactly why none of them is the default. A meeting has to tell two people apart, and engines differ in how well they do that, so it is worth trying one before committing.
Meeting summary model
The AI model used to summarize meetings. null (the default, labeled "Automatic") inherits the global Utility AI model. Override with a specific model string to route meeting summaries to a different provider. Advanced; Settings → Sessions → Per-feature AI models. Synonyms: meeting notes model.
MemPalace
Enables the MemPalace episodic memory system, which mines your session history on a background schedule and writes an .mcp.json file that gives AI sessions access to your long-term context. Off by default (opt-in). Turn on in Settings → Features once you are comfortable with Omniscio having read access to past sessions for memory extraction.
Underline menu letters
Shows one underlined letter on each item of a menu. While the menu is open, press that letter to run the item, so you can work through menus from the keyboard. Turn it off if you do not want the underlines or letter keys in menus. Find it in Settings > Keyboard. Default: on.
Mind map model
The AI model used to generate mind maps. null (the default, labeled "Automatic") inherits the global Utility AI model. Set a specific model string to override it for mind map generation only. Advanced; Settings → Sessions → Behavior → Advanced session settings → Limits, as Mind Map AI model.
Mind Map AI daily limit
The most Mind Map AI may spend in one day, in US dollars — the cap on the AI calls behind expanding a node and generating a map. When the day's spend reaches it, Mind Map AI pauses until tomorrow, and you can still edit the map by hand. Choose between $0.50 and $50; the default is $1. While sign-in is required and Omniscio cannot yet confirm your plan, the limit is held at $1 even if you set it higher; it goes back to your setting once the plan is confirmed. Find it in Settings → Sessions → Behavior → Advanced session settings → Limits, directly under Mind Map AI model. The limit counts each AI call that finishes. A call you cancel, or one that times out after 30 seconds, is not counted, even though the AI service may still charge for it, so your provider bill can run a little past this number on a day with cancels or timeouts.
Mission Control
Master gate for the Mission Control project-management integration — the isolated plugin sidebar panel and all its sub-features. Off by default (opt-in / in-development). Turning it on reveals the Mission Control sidebar row, the PM inbox integration, the calendar overlay, and related settings. Turning it off hides all three but leaves your stored configuration intact — flipping it back on restores everything.
Show text on the collapsed Omni panel
When ON (default), the slim Omni strip on the right edge of Mission Control (the Omni panel while it is collapsed) shows Omni's name (or the custom name you gave it), written sideways and centered under the Omni mark. When OFF, the strip shows only its arrow and the Omni mark, plus the attention dot when something needs you; hovering it still names it, screen readers still announce it, and clicking it still opens the panel. It is in Settings > Mission Control, next to the other Omni settings, and only appears while Mission Control is on.
Mobile device location
When on (the default), your phone reports its own coarse location — city, region, and country, resolved from its public IP by a cloud function — so the device-approval card in Omniscio can show where a device is connecting from. Turn off to disable that lookup entirely: the phone makes no geo call and the approval card shows only device and network information. This is a privacy control, not a security gate.
Mobile fast load
Makes the Omniscio mobile web app load faster on phones by loading each screen's feature code the first time you open it, instead of pre-loading all of it at boot. On by default since 2026-09-24; existing installs were switched on once, and if you turn it off it stays off. When on, the phone parses less JavaScript before the UI mounts. Turn it off to load everything up front — steadier on a patchy connection, where a screen opened for the first time after an update has to wait for its code (on a dropped connection the phone shows "Waiting for your computer…" while it waits). Takes effect on the next page refresh; no rebuild needed. An emergency env override (AMC_MOBILE_EAGER_BOOT=1) forces the eager behavior even when this is on.
Mobile font size
The text size used in the Omniscio mobile web view. Separate from the desktop font size; defaults to extra-large because phones benefit from larger text out of the box. Options: extra-small, small, medium, large, extra-large (default), xx-large, xxx-large. Change in Settings → Appearance on your phone. The header icons — including the inbox snooze and archive buttons — grow and shrink with it too, so they stay in proportion with the text.
Mobile header style
The visual style of the top header bar in the Omniscio mobile view. Options include: current (frosted glass + accent glow — the default), bare, seamless, solid, dawn, card, hairline, condensed, island, and horizon. Applies live without a restart; mobile only. Experimental — visible only when the "Mobile header and row style" lab feature is on in Settings → Appearance on a phone.
Mobile instant shell
Controls the mobile service worker cache strategy. When on (the default), the installed mobile web app serves its cached shell immediately on every open and downloads new builds in the background, swapping atomically when complete. This makes every subsequent open feel instant. Off restores the legacy network-first behavior where each open waits for the latest index.html from the server. Renderer-side only — takes effect after the next page refresh.
Let plugins open on your phone
When on (the default), a turned-on plugin whose manifest declares "mobile": true opens on your phone — and in the Phone Simulator — inside a walled-off page. That page cannot see your sign-in, your stored data or the rest of the app, and the plugin can do only what its permissions allow on this computer, minus the actions a phone always refuses. Turn it off to keep every plugin except the six built-in phone plugins on the desktop only; those six are not affected. Found in Settings → Remote Access → Devices.
Mobile reliability alerts
When on, Omniscio raises an inbox alert if it detects that the mobile web app's load watchdog is struggling — for example if the phone appears to have lost reliable connectivity to the local server. Off by default. Enable if you want proactive notification of mobile connectivity issues.
Alert me if push stops reaching my phone
On by default. Omniscio raises one inbox alert if a phone or browser is set up for push notifications, sessions need your attention, and nothing was sent to any device at all — the case where your phone has quietly stopped being notified, which otherwise nobody notices until you realise you have been missing things. The alert clears itself as soon as push reaches your devices again, and it is an inbox alert only: it is never sent as a push, because the thing that is broken is the push. It stands down on its own when you have turned Mobile Access off, so pausing access is never reported as a fault. Turn it off if you would rather not be told. Found in Settings → Sessions → Mobile.
Mobile row density
Controls how tall the session-list rows are on phones. normal (the default) uses standard row height; compact condenses rows so more sessions fit on screen. Applies live; desktop rows are unaffected. Experimental — shown in Settings → Appearance on a phone only when the "Mobile header and row style" lab feature is on.
Automatically use the newest model
When an update brings a newer model in the same family as your Claude or Codex default (for example a new Opus), Omniscio can move your default to it and tell you in your inbox. You can undo each switch from that card with one click, and the card also lets you turn this setting off.
One thing to know: while "Use Omniscio's recommended models" is on, Omniscio's own recommendation decides your default, so this setting only notes the new model and does not switch anything or send a card. It does its switching once you have picked a default model yourself, which turns the recommended models setting off. It checks when Omniscio starts. Find it in Settings > Sessions. Default: on.
Keep working when one model runs out
Some Claude models have their own usage limit that runs out before your overall allowance does. With this on, a session that hits one of those limits keeps going on your backup model instead of stopping, and it goes back to its original model after the limit resets. A short note in the session tells you when it switches.
With this off, the session stops and you get one inbox card asking whether to switch just that session. The card also offers to turn this setting on for every session. Either way, nothing switches unless your account still has room overall, and this does not apply to cloud sessions. Find it in Settings > Accounts. Default: off.
Narration mode
Controls whether and how the spoken-audio narration of AI replies fires. off disables narration entirely — no script is generated and no play button appears. click (the default) generates a spoken script and shows a per-message play button, but does not auto-play. auto generates the script and automatically plays it on each new completed reply in the open session. Requires the spoken-narration in-development feature to be revealed. Start with click to avoid unexpected audio.
Narration prompt override
An optional custom prompt that replaces the built-in prompt used to generate the spoken-narration script. Empty by default (uses the built-in prompt, which is recommended). Set only if you want to tune the wording or style of narrated recaps. Up to 10,000 characters. Part of the spoken-narration in-development feature.
Narration read view
When on (the default), displaying a message's spoken-narration script also shows a "read view" overlay — a clean text rendering of the narration script alongside the play button. The play button itself is controlled separately (by whether narration data exists). Turn off to hide the read-view overlay while keeping narration playback. See spoken-narration-contract.md.
Narration voice reply auto-send
When on, the voice-reply window opened from a narration play button automatically sends your spoken words back to the session once you pause on a sufficiently long phrase — hands-free mode. Off by default, which means you review the transcription and send manually. Part of the in-development "Reply by Voice" narration feature; requires voice input to be on.
Enable Vault (KMS)
Turns on the Vault, Omniscio's Markdown notes feature, also called KMS. It adds a Vault entry to the sidebar where you read and write notes that are stored as ordinary files in a folder on your computer, with links between notes, backlinks, tags and search across all of them. It also puts your notes into search, and makes Vault actions such as a new note, finding a note and saving a web page available in Quick Launch.
Turning this on does nothing useful until you also pick the folder that holds your notes. Do that on the same page: click Show advanced, then in The Vault group use Vault folder → Choose folder…. Turning it off hides the Vault again and deletes none of your notes; the regular cleanup of the Vault trash folder keeps running (see the privacy and safety notes on the Vault page). Find it in Settings > Features as "Enable Vault". Default: off.
KMS Vault Overview
Shows or hides the Vault Overview table in the Knowledge Management System (KMS) panel — a searchable list of every indexed note with its title, summary, and size. On by default whenever KMS is enabled. Turn off if you prefer not to see the overview table surface. The underlying indexing is unaffected.
Vault visual theme
Which of the Vault's own colour themes the note editor uses — the default look, Glassmorphism, Dark, or any custom theme you have made. It applies only inside the Vault, so the rest of Omniscio keeps the theme you chose for it. Find it in Settings > Features > The Vault > Appearance.
Vault colour mode
Whether the Vault draws its own light or dark surface: System (default) follows your computer's light/dark setting, Light and Dark pin it. Turning on a custom theme overrides this.
Vault folder
The folder on this computer that holds your Vault notes. Choose it with Choose folder…; the folder is checked first and only saved if it is a real folder that is not a system folder or a shortcut, so a bad pick shows a plain reason instead of failing quietly. Picking a different folder switches the Vault to it. Default: empty, which means no Vault is set up yet.
Keep only pinned notes as tabs
When on, opening a note replaces the unpinned tab instead of adding another one, so only notes you pin stay open. Off by default, which keeps one tab per note you open.
Vault summary daily cost cap
The most the Vault may spend in one day on AI: note summaries and the editor's AI actions. In dollars, from $0.50 to $50.00, default $2.00. When the day's spend reaches it, both pause until local midnight; nothing is lost, and the notes that were waiting are picked up on a later run. Bulk-seed spend counts against the same day. Find it in Settings > Features > The Vault > Vault summary options.
Vault bulk-seed per-run cost cap
The most one bulk-seed run — the button that summarizes every note in the vault — may spend, on top of the daily cap above. In dollars, from $1 to $50, default $5. The run stops at this cap, at the daily cap, or when you press Cancel, whichever comes first. Find it in Settings > Features > The Vault > Vault bulk-seed.
Vault image analysis cost cap
The most the Vault may spend in one day on describing pictures: the per-image title, description, text-in-the-image and tags. In dollars, from $0.50 to $20.00, default $2.00. The check runs before each batch, so a run can overshoot by at most one picture's cost. Find it in Settings > Features > The Vault.
Refresh stale summaries automatically
When on, the Vault re-summarizes notes whose text has changed since their summary was written, on a timer, without asking. Off by default — summaries are then only written when you ask for one or start a bulk-seed. Find it in Settings > Features > The Vault > Vault summary options.
How often to refresh a summary
How long the automatic summary refresh waits between passes over the vault, in minutes. From 1 minute to 1440 (a day), default 5 minutes. Shown only when automatic refresh is on; a shorter interval means fresher summaries and more AI spend.
Wait before refreshing the same note again
The shortest gap between two automatic re-summaries of the same note, in minutes. From 1 minute to 1440 (a day), default 15 minutes. It stops one note you keep editing from being re-summarized on every pass.
Expose Vault to spawned agents
This switch no longer changes anything. The per-session Vault tools it used to switch on were retired, so turning it on gives a session no new way into your notes and turning it off keeps no session out. Agents reach your notes through the app's own Vault routes instead, and a session you start from the Vault reads and edits its Markdown files directly. The row is still there so an existing setting keeps working, and it stays off by default. Find it in Settings > Features > The Vault.
Vault editor font
The typeface the note editor uses: Sans (default), Serif, or Mono. Only the editor changes; the rest of Omniscio keeps its own font. Find it in Settings > Features > The Vault > Appearance.
Vault editor text size
How large text reads in the note editor: Small, Medium (default), Large or Extra large. It also scales the note's headings, so the shape of the page stays the same at every size.
Vault editor line width
How wide the column of text is: Narrow for short lines that are easy to scan, Normal (default), Wide, or Full for the whole pane. Narrower columns read better on a wide screen.
Vault editor page colour
The paper behind your notes: Theme (default) follows the app's own background, Paper is plain white, and Sepia is a warm off-white that is easier on the eyes for long reading.
Vault sidebar side
Which side of the window the note list sits on: Left (default) or Right.
AI Knowledge Digest
Periodically reads your sessions, pulls out the notable ones and writes a curated markdown digest into your vault's Agent Insights folder, one note per project. Off by default. It needs the Vault switched on as well, and switching it on here takes effect the next time Omniscio starts. Paid like the other built-in AI helpers. Find it in Settings > Features > The Vault.
Daily digest
Writes one digest every day at the hour below. Off by default — the weekly digest is the one that ships on.
Run the daily digest at
The hour of the day the daily digest runs, in your local time, from 0 to 23. Default 6, which is 6:00 AM.
Weekly digest
Writes one digest a week, on the day and hour below. On by default.
Run the weekly digest on
The day the weekly digest runs, 0 = Sunday through 6 = Saturday, in your local time. Default 1, which is Monday. The hour beside it defaults to 6, which is 6:00 AM.
AI Knowledge Digest per-run cost cap
The most one digest run may spend on AI, in dollars from $0.01 to $10.00, default $0.10. The run stops when it reaches the cap, so a busy week costs the same as a quiet one.
Deep-dive session limit
How many of the most interesting sessions get the expensive full-transcript read on each run, from 1 to 50, default 10. Every other session is still summarised from its lighter signals; this only bounds the costly pass.
Digest sections
Which parts appear in a generated digest note: Key decisions, Patterns and themes, Bugs and fixes, Lessons learned, Session activity summary and Vault activity are on by default; Open questions is off, because it tends to be noisy. Turn any of them off to shorten the notes.
Session finished sound
Controls the chime that plays when a session finishes its turn and nothing is waiting on you. It is one row in the list of which events make a sound, next to Needs You, Approval waiting, Stalled and Error, and each row has its own on/off switch and sound picker. Turning this one off silences only the finished chime. The others keep working.
Like the other chimes, it stays quiet during your quiet hours. Find it in Settings > Notifications, under Which events make a sound. Default: on.
OpenAI admin key (vendor key watch)
An optional OpenAI admin key the vendor key watch uses to read OpenAI's cost report for the API key Omniscio holds. OpenAI only reports spend to an admin key; create a read-only one if your organisation offers it at platform.openai.com → Settings → Admin keys. Stored encrypted, kept in the main process, used for reads only, and never shown anywhere. Leave it empty and OpenAI shows as "not watched — needs an admin key". Secret text; accepted values: string (≤500).
OpenCode model
The model OpenCode sessions use, specified as provider/model (for example openrouter/moonshotai/kimi-k2). null or empty (the default) uses OpenCode's own built-in default model with the OpenCode/Anthropic key. The provider prefix selects which stored credential Omniscio injects at session spawn. Change this in Accounts & Providers → OpenCode if you want a specific model for all OpenCode sessions.
Who pays for each provider
Replaced by Who pays & who serves. The "Use Omniscio … credits" switches that wrote this setting are gone: who pays is now a row in each model family's supply list, and your earlier choices were carried into those lists. What follows describes it as it was.
Your choice, for each AI provider separately, between paying with your own key for that vendor and paying with Omniscio credits. Setting one provider never changes another — you can run DeepSeek on your own key and buy GLM on credits at the same time.
Each provider you have chosen for is stored as a ranked list of funding sources: own-key (your own vendor key, billed by that vendor directly) and omniscio (Omniscio's gateway key, which draws down your prepaid credit balance and stops at zero). The session uses the first source that is actually ready, and moves to the next if the active one runs out mid-session.
A provider you have never chosen for is absent from the list, which means your own key first — nothing starts spending credits by surprise. Ranking is only a preference: each source still has to pass its own readiness and plan checks, and an unready one is skipped rather than forced.
Replaced eight older settings on 2026-09-13 — a single global funding order plus seven per-vendor "use Omniscio credits" toggles — which between them could not express a per-provider choice. Every existing opt-in was converted automatically. Object mapping each provider id to its ranked list; accepted source values: own-key | omniscio.
Outbound webhooks
Master gate for the outbound webhook dispatcher — the feature that delivers a curated set of Omniscio events to customer-configured callback URLs. Off by default (in-development). The dispatcher re-reads this flag on every event so toggling it takes effect live without a restart. The feature is inert until the webhook management UI is available; turning this on early is safe but has no visible effect until webhooks are configured.
Overseer groups
The named groups you have made for your Overseers — folders in the Overseers section of the sidebar and on the phone. Membership is by Overseer, not by position, so an Overseer keeps its group across the restarts that give it a new session; a group's own position in the list is kept with the rest of the arrangement, and its collapsed-or-expanded state is stored on the group so the phone and the desktop agree. Empty by default (up to 50 groups, names up to 80 characters). An Overseer placed in two groups belongs to the first; a group naming an Overseer that no longer exists is ignored. Arrange them in the Overseers section.
Paste rich text as Markdown
When on (the default), pasting rich text (HTML from a webpage, a Google Doc, etc.) into an Omniscio composer converts it to Markdown before inserting it. Turn off if you want rich-text pastes stripped to plain text instead. Note: content pasted from Google Sheets is always stripped to plain text regardless of this setting, because the rich-text representation is unreadable noise.
Hand agent messages over mid-turn
How a message from another agent is delivered to a session that is already mid-turn. On by default: the message is handed over at the session's next tool boundary instead of waiting for its turn to end — measured waits of 16, 47 and 56 minutes are why this exists. Only agent traffic is carried this way; your own queued typing still arrives as a new turn. Off restores the earlier behaviour, where every held agent message waits for the turn to end. Read at every hand-over, so a flip applies at the next tool boundary with no restart. The AMC_DISABLE_PEER_TOOL_BOUNDARY environment kill switch outranks this setting. Toggle; accepted values: true | false.
Skip redundant read-permission requests
Adds a short note to each session telling the agent that reading files is already approved here, so it should just read the file instead of stopping to ask for read permission it already has. This cuts out pauses that look like they need a click but do not. If a read is truly blocked, the agent still gets a clear error and shows it. The note is left out of the built-in Ask Omniscio helper sessions. It only changes what the agent is told, not what it is allowed to do. Find it in Settings > Features. Default: on.
Agent persona
Controls the persona block injected at the top of every session's system prompt. When on (the default), the agent adopts an upbeat, proactive, resourceful, friendly persona that defaults to yes, takes initiative, assumes a request is possible, and drives work to a finished result. Turn off to remove the persona and character block entirely, leaving a neutral tone. Settings → Sessions → Prompting.
Pi model
The model Pi sessions use, specified as provider/model (for example anthropic/claude-sonnet-4-5 or openai/gpt-4o). null or empty (the default) uses the built-in Pi default model. The provider prefix selects which stored credential is injected at spawn. Change this in Accounts & Providers → Pi.
Session tool picker tools
A list of tool/engine IDs that appear in the new-session Tool dropdown (the start-config picker). Unset by default, which shows all available engines. Providing a list trims the picker to only those IDs. claude-code is always shown regardless. This only affects the per-session picker — not the per-project default setting.
Pika voice
The voice preset sent to the Pika Speech text-to-speech service. Accepts a named preset string (one of Pika's PIKA_PRESET_VOICES). Default is calm_documentary_narrator. Pika is user-key-only (no relay); change this in Settings → Voice Control if you have a Pika key and want a different voice.
Plugin auto-check updates
When on (the default), Omniscio keeps your installed plugins up to date. At launch and on a background interval it applies an update that asks for nothing new by itself, and when a newer version wants access you have not approved it stops and asks you first, with the permission list shown. Turn off to disable the checks entirely: nothing is checked and nothing is updated in the background.
Lock down plugins' AI agents
When on (the default), every new AI agent session a plugin starts runs on a locked-down floor. It gets none of your own setup: no instructions, memory, skills, hooks, saved permission rules or add-on tool servers. It works in its own folder outside the app's data, uses only the tools its job needs (never more than the plugin was granted when you installed it), and runs only on Claude-family engines, on this computer.
- A hidden background session (such as a Decks build) answers in text only, unless its plugin asks for web search and public web pages. It never asks you anything: a call outside its floor is simply refused.
- A visible session (one you can open and answer, such as a Forms routed response or a GitHub Integration fix) asks you before every command and every web request, and can never commit or push to main or master.
- Every plugin session on the floor is refused outright from the app's keys, the app's data folder, your own Claude folder and your sensitive files, even when your own trust level is Full trust.
Why it exists: a plugin hands its agent content from outside, such as a form answer, a web page, an email or an issue. A hidden instruction in that content could otherwise steer an agent running with your full access.
Turning this off is the emergency rollback: every plugin's agents get your full setup and access again from their next launch. To trust just one plugin, use that plugin's own full-access switch instead. A session a plugin locked on purpose (an AI that can only talk) stays locked either way. Sessions created before the floor existed are not affected. The toggle is in Settings → Plugins, just above Developer Mode, and has no keyboard shortcut. You or an agent can change it, like any other setting; a plugin cannot.
Plugins whose AI agents have your full access
The plugins whose AI agents you have given your own full access, one plugin each. Empty by default, so every plugin's agents run on the locked-down floor described above. A plugin listed here starts its agent sessions with your own setup, your trust level and any engine, as before the floor; every other plugin stays locked down.
Not recommended: a plugin's agents read outside content that could steer them. The change takes effect the next time each of that plugin's agent sessions starts. A session the plugin locked on purpose stays locked. Set it with the "Give this plugin's AI agents my full access (not recommended)" switch: open the Marketplace, find the installed and enabled plugin's card, and expand Settings on it. The switch has no keyboard shortcut. You or an agent can change it, like any other setting; a plugin cannot.
Background AI sessions a day
How many AI sessions one app may start in the background each day, set separately for each app. Every background session an app starts counts — a build it runs for itself, research it does before writing something, a run it kicks off on a schedule. Those sessions never appear in your session list.
An app you have not changed gets 20 a day, which is the number its consent sentence states. Raise it for one app and only that app's background work may go past 20; lower it to rein that app in, or set it to 0 and that app starts no background sessions at all. The limit covers a rolling 24 hours rather than a calendar day, and it is enforced in the app itself, whichever way the app asks for the session — so the number on the card is the number in force.
The ceiling is yours, not the app's: it is stored with your own settings, never in the app's own settings, so no app can raise it. Change it in Settings → Plugins, on the installed app's card, under "Background AI sessions a day"; it applies from that app's next background session. You or an agent can change it, like any other setting; a plugin cannot.
Which boards appear on the calendar
Narrows the Mission Control calendar overlay — due dates and sprint timelines drawn on your calendar — to specific projects. Leave it empty — the default — and every project shows up. Change it from the gear at the right edge of the Calendar panel's tab row: the "Projects" checklist under "Mission Control due dates", where a checked project is shown. It is no longer on the Settings → Mission Control page; searching Settings for it opens that Calendar control. Has no effect unless the PM calendar overlay itself is on. Holds up to 100 projects.
PM calendar overlay
Shows Mission Control project due dates and sprint timelines as an overlay on the calendar view. On by default when Mission Control is enabled. Turn it off from the gear at the right edge of the Calendar panel's tab row — the "Show due dates" switch under "Mission Control due dates" — if you want the calendar to show only your own events. It is no longer on the Settings → Mission Control page; searching Settings for it opens that Calendar control. The same block lets you limit the overlay to specific projects.
Per-board inbox preferences
Retired. This used to hold one inbox choice per board (All / Assigned only / None) from dropdowns in Settings → Mission Control. Those dropdowns are gone: each saved choice was moved once into Mission Control's own per-board list — a board set to None stays muted, and a board set to Assigned only behaves as Involves me — and the setting is then emptied. How much reaches your inbox is now set by the Inbox level.
Mission Control Inbox level
How much Mission Control activity reaches your Omniscio Inbox, set in Mission Control → Settings → Updates: Everything (every change except your own), Involves me (things assigned to you, that mention you, that you follow, or on a board only you use), Mentions me, or Nothing. Until you pick one, it is Everything if your setup answers say you run a department or you chose the Operations style, and Involves me for everyone else. If you already used Omniscio before this setting existed, it starts on Everything (even if Mission Control was switched off then) so nothing you were getting disappears. Due and overdue rows never go wider than Involves me, at any level. Changes kept out of the Inbox still show in What Changed, and the Mission Control project events in Inbox switch still turns everything off.
Tell me when an automation fails
When on (the default), a Mission Control automation that fails puts a row in your Inbox. Turn it off if you would rather check automation runs yourself. Set in Mission Control → Settings → Updates.
Tell me when an automation changes something that involves me
When on (the default), an automation that changes an item assigned to you, an item you follow, or anything on a board only you use puts a row in your Inbox. Automation runs on other people's work only do when your level is Everything. Set in Mission Control → Settings → Updates.
Ask before opening Omni
When you ask Mission Control a question it can't answer on the spot, it shows what it can see right away and a "Keep going with Omni" button (both show the custom name if you gave Omni one). On by default, because an Omni session is a paid AI session: one only starts when you tap the button, and it opens in the sessions sidebar with your question and project already loaded while you stay on the dashboard. Turn it off to have the session open on its own. Questions Mission Control can answer itself (status, blocking, overdue, attention, summary and risk) never open a session either way.
PM inbox digest mode
When on, Mission Control stops popping up a desktop notification (and its sound) for each project change — assignments and approaching due dates are collected in your daily digest instead. The changes still appear as rows in your inbox, and overdue items still notify immediately. Off by default, so each change pops up as it happens. Turn on if you prefer a periodic summary over a continuous stream of Mission Control notifications.
PM inbox
Enables the project-management inbox integration, which surfaces PM board events (assigned tasks, status changes, approaching due dates, overdue items) as inbox rows. On by default when Mission Control is enabled. Turn off to stop PM events from appearing in your inbox while keeping the rest of Mission Control active. Off also stops the "Mission Control changes on your board" cards and their desktop pop-ups; those changes are marked as handled, so turning the switch back on does not replay them. You can also turn it off from the inbox itself: open a Mission Control item and press Turn off these alerts, then confirm. The toast that follows has an Open Mission Control button that brings you back here to turn them on again.
PM inbox quiet hours
When on, Mission Control's desktop notification and its sound are held during a configurable daily time window so you are not disturbed outside working hours. Changes still appear as rows in your inbox, and overdue items still notify immediately. Off by default. Set the window with pmInboxQuietHoursStart and pmInboxQuietHoursEnd (in HH:MM format) and the time zone with pmInboxQuietHoursTimeZone.
PM inbox quiet hours end
The time (in HH:MM 24-hour format) when the PM inbox quiet-hours window ends and notifications resume. Default 07:00. Interpreted in the time zone set by pmInboxQuietHoursTimeZone. Only active when pmInboxQuietHoursEnabled is on.
PM inbox quiet hours start
The time (in HH:MM 24-hour format) when the PM inbox quiet-hours window begins and notifications are suppressed. Default 22:00. Interpreted in the time zone set by pmInboxQuietHoursTimeZone. Only active when pmInboxQuietHoursEnabled is on.
PM inbox quiet hours time zone
The time zone your PM inbox quiet hours are measured in — the daily no-notification window for project-management items. Leave it unset — the default — and quiet hours follow your computer's current zone. Set a specific zone (an IANA name like America/New_York) to pin them, so quiet hours stay at the same local wall-clock time even if you travel or your machine's zone changes. Only relevant while quiet hours are turned on.
Omni's memory of you
On by default. When on, Omni keeps facts it learns about you from your conversations — how you work, what you care about — so it can help you better. They are stored only on this computer, and you can read and delete them in the Memory card of the Omni area. Turn it off and Omni stops keeping new facts about you this way; your Mission Control projects and tasks are unaffected. Find it in Settings → Mission Control, in Omni's memory card.
Send nudges right away
Controls what the Nudge button on a quiet item does. Off (the default): Nudge opens the item's updates with a ready-to-send message to the owner that you can edit before it goes. On: Nudge posts that message straight away and shows "Nudge sent". Only shown while the quiet-item flag is on.
PM setup preset
The project-management workflow preset you selected during Mission Control setup. Options include general presets (tasks, projects, operations, first-agent-board) and tool-specific ones (trello, jira, asana, linear, notion, clickup, airtable, engineering-linear, engineering-jira). Written by the setup wizard; change it later in Mission Control → Settings → How you work (Change), or redo the setup checklist from the same place.
AI comfort level
Records how much AI help you want from Mission Control, from the answer you gave during setup. Low keeps AI off: Mission Control never sends anything to an AI, so its AI answers are off. Medium and High turn AI on, and High uses it more often. Left unset, it reads as Low. Change it in Settings → Mission Control, or during setup on the AI Comfort step. Accepted values: low, medium, high; default: unset, which reads as low.
Add my out-of-office from Google Calendar
On by default. With Google connected, Mission Control reads your own primary Google Calendar on this computer every 15 minutes and adds your out-of-office, PTO, vacation, holiday and leave days to Who's out, where your team sees them. Only the dates are kept, never the event's title or details. Turn it off and this computer stops reading your calendar; days already added stay until someone removes them in Who's out. The switch is in Statistics → Project Analytics → Busy seasons and risk → Who's out. See Mission Control cross-board intelligence.
Ask automatically when I open the home page
Part of the "What should I do next?" card on the Mission Control home page, which is still in development. By default the card asks the AI only when you press its button. Turn this on and Mission Control asks for a fresh answer every time you open the home page. It reaches the home page only: the What should I work on next? widget on an Omni control center page asks only when you press its button, whatever this is set to.
Every ask runs an AI request, so when you turn this on Omniscio asks you to confirm first. Turning it off needs no confirmation. The switch only shows up in Settings > Mission Control while the What Next feature is available to you. Default: off.
Use a faster, cheaper AI for What Next
Off by default. Mission Control's "What next?" answer runs on the smart AI tier, because it was the only one that stated no false facts when tested against planted trap boards. It runs on your own AI key, so each answer costs about 4 cents. Turn this on to answer on the worker tier instead: faster, roughly half the cost, but less careful. "Look deeper" always uses the most capable tier either way. See Mission Control cross-board intelligence.
Pomodoro — auto-enable Focus Mode
When on (the default), starting a Pomodoro focus block automatically enables Focus Mode, and entering a break automatically disables it. If you manually toggle Focus Mode during a run, you take back control and the Pomodoro service stops managing it for the rest of that run. Turn off if you want to run the Pomodoro timer without it touching your Focus Mode state.
Pomodoro — bring to foreground on phase change
When on, the Omniscio window forces itself to the foreground every time the Pomodoro timer transitions between focus and break phases. Off by default — the audible chime and tray badge are usually enough. A per-preset bringToForegroundOverride takes precedence, and Focus Mode can still gate this behavior.
Pomodoro schedule
Master toggle for the Pomodoro schedule. When off, all scheduled Pomodoro session rows are ignored — no pre-start toasts and no auto-starts. On by default so a newly created schedule fires as expected. Turn off to pause the schedule temporarily without deleting it.
Post-send navigation
An ordered list of actions Omniscio tries in sequence after you send a message from the inbox. Options are stay (keep the current session open), next_in_project (move to the next session in the same project), next_in_inbox (move to the next inbox item), and waiting_room (go to the waiting room). The default order is next_in_project, next_in_inbox, stay — Omniscio picks the first option that has a valid destination. Reorder or remove options in Settings → Inbox.
Scope — typed replies only. This setting governs what happens after you type and send a reply. One-click triage gestures deliberately ignore it: archive, snooze, Continue, and Allow/Deny on a command prompt all advance the way archiving does — to the next session that needs you, and failing that the next visible session in the area you are already in. That is why approving a command can move you somewhere this list does not name.
Skip inbox cards for my own pull requests
When a pull request you wrote lands on this computer's master, Omniscio normally leaves a short landed card in your inbox. With this on, your own requests are skipped: no card is made for them, so the cards you get are the ones somebody else wrote. A request counts as yours when the person who opened it is the account signed into this install — the same identity the contributor report uses, so committing from another machine or another address still counts as you. If Omniscio cannot tell who the author is, the card is delivered exactly as before. Your pull requests still land normally; only the card is skipped. The card for a request nobody opened, and every other developer's card, are untouched. On by default; turn it off if you want to see your own lands too. You find it in Settings > Inbox.
Default: on.
PR Merge Queue — auto-merge poll interval
How often (in minutes) the background queue refresh re-fetches every configured repo — and, because one timer drives both, how often the auto-merge daemon ticks when it is on. Default is 10 minutes; valid range is 1–120. Values outside the range are clamped at startup. Relevant whenever prMergeQueueEnabled is on with at least one repo configured; it is independent of prMergeQueueAutoMergeEnabled, so the saved snapshots and the Inbox counts stay fresh even with auto-merge off. Changing it takes effect on the next Omniscio start — the running task keeps its boot-time interval.
PR Merge Queue
Master gate for the PR Merge Queue integration. When on, the sidebar virtual project and the toolbar icon become visible, and the background poll runs. When off, both of those are hidden and the background poll stops, but your stored repo configuration and auto-merge daemon settings are preserved — toggling back on restores everything. The Settings → PR Merge Queue section is always in the Settings nav and is not gated by this flag, so finding it there does not mean the feature is on. Off by default; existing installs that already have repos configured are auto-enabled once via a one-time migration. This switch is now the PR Merge Queue Install in Marketplace → Inbuilt rather than a Settings row.
PR Merge Queue — shard count
The total number of machines sharing the PR merge workload (the modulo divisor used when sharding is on). Default is 2. A value less than 2 makes sharding a no-op so this machine handles everything. Only meaningful when prMergeQueueShardEnabled is on. Set each machine to a distinct shard index and the same shard count so they partition the work without overlap.
PR Merge Queue — sharding
When on, the auto-merge daemon only auto-processes pull requests whose PR number modulo the shard count equals this machine's shard index. This lets two or more independent Omniscio installs share the same repo's queue without ever merging the same PR twice. Off by default. Manual merge actions (the Merge button, MERGE_PR, MERGE_BATCH) are never shard-filtered — an explicit action always works. Sharding applies to the in-app auto-merge daemon and to the PR Janitor's analysis run, which reads the shard the daemon exports and judges only this machine's PRs; nothing else consults the shard config.
PR Merge Queue — shard index
This machine's 0-based position in the shard split (0 ≤ index < count). Give each install a distinct index — the first machine takes 0, the next takes 1, and so on. Default is 0. An out-of-range or negative value is fail-safe: this machine auto-spawns nothing rather than risk a double-merge.
Let agents restart your programs so they survive quitting
Normally a program an agent starts for you ends when Omniscio quits. This setting lets an agent restart a program you were already running in a way that keeps it running after Omniscio closes, until you close it yourself or restart the computer. It only works for a program that was already running before the session started, never a shell or script host.
While it is off, the first time an agent asks for this, an inbox card appears. You can allow it for that one session or allow it for good, and choosing "for good" turns this setting on. Turning it on here skips the card for every session. Find it in Settings > CLI Control. Default: off.
When I switch projects
Decides what opens when you switch to a project from the sidebar. Needs You, else a new chat (the default) opens that project's top Needs You row — the session, alert or approval waiting on you there — and otherwise drops you straight into a new, empty chat that is ready to type in. The chat I was last on restores the chat you were last viewing in that project, falling back to its most recently active session. Either way, coming back with a Back gesture, or the app reopening the project you had open last, still returns you to exactly where you were: this setting only governs opening a project deliberately. Found in Settings → Sessions → Behavior.
Provider config sync targets
The list of other AI tool IDs (for example codex, gemini, opencode, cursor) that Sync to Other AI Tools will install your skills, and the instructions and MCP servers you opted into, into. Empty by default, and added on top of the tools you have enabled while "Sync to every enabled AI tool" is on; it is the whole list once that switch is off. Only servers you flagged for syncing are registered, and a server that carries a credential is never written to another tool. Found in Settings → Automations & Rules → Sync to Other AI Tools (desktop only).
Quick Chat per-target hotkeys
Master enable for global keyboard shortcuts assigned to individual Team Chat Message targets (people and channels). Off by default. When on, each saved target that has a hotkey configured registers it as an OS-wide shortcut that opens the Quick Launch composer pre-aimed at that person or channel. Turn on only if you want one-key access to specific teammates or channels from anywhere on your machine.
Quick Email saved recipients
The list of contacts saved for Quick Email — the fast one-key email composer in Quick Launch (Ctrl+Space). Each entry has a display name, an email address, and an optional per-recipient global hotkey. Up to 50 recipients. Add and manage them in Settings → Quick Email. When a hotkey is set for a recipient, pressing it from anywhere on your machine opens the Quick Launch composer pre-filled for that person.
Quick Email signature
An optional signature appended to the body of every email sent from the Quick Launch Quick Email composer. Blank by default (nothing appended). This is specific to Quick Email — it does not pull from your Gmail web signature. Up to 1,000 characters. Set it in Settings → Quick Email.
Quick Email undo window
How many seconds a sent Quick Email is held in an undoable pending state before it is actually delivered. During this window you can cancel the send. Default is 10 seconds; valid range is 1–30. Increase if you frequently catch mistakes after hitting send; decrease if you want delivery to be more immediate.
Quick Launch — follow cursor
When on, the Quick Launch floating modal opens on whichever monitor your mouse cursor is on rather than always opening on the primary display. Off by default. Useful if you work across multiple monitors and want Quick Launch to appear where you are looking. Takes effect on the next Ctrl+Space press — no restart required.
Quick Launch hotkey
The global keyboard shortcut that opens the Quick Launch floating modal (default Ctrl+Space; on macOS the default is Cmd+Option+L, since Cmd+Space is Spotlight). Accepts a single Electron accelerator string or a list of alternates. Registered with the OS only when quickLaunchHotkeyEnabled is on. Change it in Settings → Hotkeys if it conflicts with another application (note: on Windows, Ctrl+Space is also the IME input-method toggle, which can cause a conflict — Omniscio will warn you if a CJK input method is detected).
Quick Launch — preserve draft
When on (the default), closing the Quick Launch composer without sending keeps your unsent draft — text and any pasted chips — so the next time you open Quick Launch it resumes where you left off. The Reset button or turning this off clears the draft instead. Sending always clears the draft regardless of this setting. Draft state is held in memory only; a full app restart always starts blank.
Quick Launch — show engine picker
When on, the New Session tab in Quick Launch shows two compact chips — Engine and Model — next to the project chip, letting you change the engine or model for just that one launch. Off by default. The resting state shows the project's resolved default; picking a chip overrides it for the next session only. Turn on if you frequently launch sessions with a non-default engine from Quick Launch.
Quiet hours
Silences the notification chime during a window you choose, for example overnight. Pick the start time, the end time and the days it applies to. The window starts at 10:00 PM and ends at 7:00 AM, every day, until you change it.
It also holds back outside-the-app alerts such as phone calls, texts and emails that Omniscio would send you, other than the ones marked critical, and the phone reminders for your tasks. Your inbox, in-app alerts and desktop pop-ups keep working. Below the switch, "Quiet for now" silences the chime for 1, 2 or 4 hours with one tap, without changing your schedule. Find it in Settings > Notifications. Default: off.
Source checkout folder
The folder of your Omniscio source checkout, used when you run the installed Omniscio rather than one you started with npm run dev. An installed Omniscio lives inside its own app bundle, which is not a checkout, so it has no way to find the background jobs that keep this computer healthy — the scheduled maintenance jobs and the broker that answers your test gates. Point this at your checkout and those jobs run from it. Leave it empty and Omniscio keeps using whatever checkout it can already find (which is what a development machine does). Set it in the Dev Pipeline panel → Setup → the "Source checkout" card. If it is left empty on a machine with no checkout anywhere, a test gate refuses by name and tells you to set this.
Board density
Controls how tightly items are packed on Mission Control boards. Dense fits the most on screen, then Compact (the default), Comfortable, and Spacious, which gives each card the most room. Pick the one that reads best at your window size. Set in Mission Control → Settings → Look; Mission Control's theme always follows Omniscio's.
Count my week as business days only
Off by default. When on, the Mission Control Risk Heatmap counts only Monday to Friday: "Next 7 days" becomes "Next 5 business days", "Next 30 days" becomes "Next 20 business days", and a project's pace is judged per weekday, so weekends never count as working time. Turn it on if your team works weekdays only. When the Mission Control company holidays switch is also on, each project's company holidays are skipped too, so they never count as working days.
Name your assistant
The name Omni, your AI project manager, uses for itself, everywhere it talks: Mission Control help, setup, the Tasks triage screens, its sessions and its replies. It also names the settings labels, the Stats tab, the Inbox filter, the keyboard shortcuts, the daily-limit alert and the Mission Control team-chat bot. The default is Omni. Type a new name and click away to save it; it shows everywhere at once, with no restart. Only the sidebar item keeps the name "AI Project Manager". Letters, numbers and spaces only, up to 40 characters; anything else is turned into spaces, and an empty box goes back to Omni. Renaming changes only Omni: the voice Session Briefing and every other assistant keep the name Omni. Found in Settings → Mission Control while Mission Control is turned on.
Match my Omniscio theme
Mission Control wears its own cyan by default. Turn this on and every part of Mission Control uses your Omniscio theme color instead: the boards, the sidebar, popped-out windows, the setup checklist and Customize windows, and the Mission Control screens elsewhere in the app. Small colored text, like links and "+ Add" buttons, is stepped darker or lighter where needed so it stays easy to read. The Mission Control logo keeps its own colors either way. Find it on Mission Control's own Settings page, in the Look card under Board density. It is off by default, and changing it takes effect at once.
Let Omni read your documents, spreadsheets and forms
Decides whether Omni, your AI project manager, may read your documents, spreadsheets and forms. Omni reads a page you open (for a form, its answers) and files what needs doing into a project you choose. It never changes your pages. The first time, Omni asks on the page itself. Turn the switch on to say yes (the setting is "allowed") or off to say no ("declined"). Until you answer, it shows as off. Found in Settings → Mission Control, under Name your project manager, while Omni is turned on.
Sync to every enabled AI tool
On by default — everything you share (your skills, plus the instructions and MCP servers you opted into) goes to every AI tool you have turned on, so nothing needs ticking by hand. Turn this off to pick the tools yourself in the target list below. Found in Settings → Automations & Rules → Sync to Other AI Tools (desktop only).
Run this many PR sessions at once
How many pull-request helper sessions run at the same time. It belongs to the Pull-request helper in Manage Sessions > Automation, which appears once Developer Guardrails is turned on in Settings > Features. The helper keeps this many sessions going and starts a new one when one finishes. Sessions already running are never cut off.
It only matters if you run Omniscio's own pull-request helper from a source copy of the app. An installed copy does not run that helper, so changing this there has no effect. Choose a whole number from 1 to 100. Default: 10.
Pull requests per session
How many pull requests the Pull-request helper gives one session to work through. Larger batches finish the whole list sooner, because each session reads its background once and then works through several pull requests. A pull request that is held for review is always worked on by itself.
It sits next to "Run this many PR sessions at once" in Manage Sessions > Automation, and the same note applies: only a source copy of the app runs the helper. Choose a whole number from 1 to 50. Default: 7.
Move a hub to another workspace
Shows "Move to workspace" on a hub's right-click menu, so a hub can be moved from your personal space into a company workspace (or back) without losing its sessions, history, files or pins. Turning this off hides the action and refuses the move. On by default.